Skip to contentSkip to conversation

2.6.9 & 11 hang on ac/dc change

DISCUSSION William D Waddington 2 participants 3 messages May 23, 2005 – Jun 15, 2005

lore
Share & tools
lore
#0 lore raw

In the course of trying to figure out why /sys/.../cpu0 isn't populated when in APM mode, I discovered that the laptop hangs on the first (?) attempt to write to the drive after a change of power source.

This is a ThinkPad T43. I see the problem with both 2.6.9-1.667 and 2.6.11-1.14_FC3 (both FC3). Doesn't seem to happen in ACPI mode. I need APM since that suspends and hibernates will on TPs. (Including this one)

It always leads to a filesystem check on reboot, and once it damaged the partition so badly that it wouldn't reboot and "linux rescue" from the FC3 DVD couldn't mount it. That time, I got "journalling error" messages on the console, and "abnormal status 0xD0 on port 0x1F7" when I tried to shut down.

I can't write to the HD to try to capture anything, so I don't have much diagnostic info. I have posted the APM boot messages, FWIW:
http://www.beezmo.com/scratch/t43_fc3/dmesg.apm

Any place I should poke to track this one down? If this doesn't belong on lkml, please point me appropriately.

Thanks,
Bill (not subscribed but lurking on fa.linux.kernel)

#1 lore raw

Re: 2.6.9 & 11 hang on ac/dc change with T43

I am having this same exact problem! I'm running 2.6.11.12 on my ThinkPad T43 which I just got yesterday. I discovered last night then when I change the power state in any way, it'll give these ATA abnormal status 0xD0 on port 0x1F7 errors.

So I began doing some sleuthing.. Bill said that this only happens with APM, so I tried it without APM. First I booted with noapm, but I found that my system was still trying to use the apm kernel module when apmd started. So, I removed apm from the kernel and uninstalled apmd. An interesting side effect that I noted here was that once apmd was uninstalled it generated some sort of power-change event that triggered this bug.

So I booted with a new kernel without any APM support and tried changing the power state -- no luck. Same errors. Fortunately I haven't experienced severe filesystem corruption quite yet, but I have a feeling that if I keep experimenting long enough, it will be inevitable.

So this is a pretty grave bug. Everytime I want to unplug my laptop or plug it back in, I have to shut it down first. I'm not sure what kernel version this bug was introduced in -- I might try building 2.6.8 and see how that works.

I believe that there are two more things I can try; Bill mentioned this problem not happening in ACPI mode. I thought that by this he meant that the problem disappeared in absence of APM, but perhaps this problem might disappear in presence of ACPI. Well, I'll try this, but I'd prefer to have working APM on this machine for suspend.

Maybe disabling the BIOS power-saving mechanisms and solely relying on Linux kernel to handle CPU throttling and such might do it, but I have my doubts.

Has anyone else had this problem? Anyone look into it? It seems that Bill posted this a while ago and has received no reply on the list.

Quoted 21 lines · William D Waddington

In the course of trying to figure out why /sys/.../cpu0 isn't populated when in APM mode, I discovered that the laptop hangs on the first (?) attempt to write to the drive after a change of power source.

This is a ThinkPad T43. I see the problem with both 2.6.9-1.667 and 2.6.11-1.14_FC3 (both FC3). Doesn't seem to happen in ACPI mode. I need APM since that suspends and hibernates will on TPs. (Including this one)

It always leads to a filesystem check on reboot, and once it damaged the partition so badly that it wouldn't reboot and "linux rescue" from the FC3 DVD couldn't mount it. That time, I got "journalling error" messages on the console, and "abnormal status 0xD0 on port 0x1F7" when I tried to shut down.

I can't write to the HD to try to capture anything, so I don't have much diagnostic info. I have posted the APM boot messages, FWIW:
http://www.beezmo.com/scratch/t43_fc3/dmesg.apm

Any place I should poke to track this one down? If this doesn't belong on lkml, please point me appropriately.

Signature · 3 lines
Misha Nasledov
[email protected]
GPG: A063 B99A 2BD3 2D48 F2D7 8E68 0F27 4D21 948F 8F06
#2 lore raw

Re: 2.6.9 & 11 hang on ac/dc change with T43

So I've done some further experiments and apparently this bug goes away when ACPI is enabled in the kernel. This is a temporary fix though..

Quoted 35 lines · Misha Nasledov

I am having this same exact problem! I'm running 2.6.11.12 on my ThinkPad T43 which I just got yesterday. I discovered last night then when I change the power state in any way, it'll give these ATA abnormal status 0xD0 on port 0x1F7 errors.

So I began doing some sleuthing.. Bill said that this only happens with APM, so I tried it without APM. First I booted with noapm, but I found that my system was still trying to use the apm kernel module when apmd started. So, I removed apm from the kernel and uninstalled apmd. An interesting side effect that I noted here was that once apmd was uninstalled it generated some sort of power-change event that triggered this bug.

So I booted with a new kernel without any APM support and tried changing the power state -- no luck. Same errors. Fortunately I haven't experienced severe filesystem corruption quite yet, but I have a feeling that if I keep experimenting long enough, it will be inevitable.

So this is a pretty grave bug. Everytime I want to unplug my laptop or plug it back in, I have to shut it down first. I'm not sure what kernel version this bug was introduced in -- I might try building 2.6.8 and see how that works.

I believe that there are two more things I can try; Bill mentioned this problem not happening in ACPI mode. I thought that by this he meant that the problem disappeared in absence of APM, but perhaps this problem might disappear in presence of ACPI. Well, I'll try this, but I'd prefer to have working APM on this machine for suspend.

Maybe disabling the BIOS power-saving mechanisms and solely relying on Linux kernel to handle CPU throttling and such might do it, but I have my doubts.

Has anyone else had this problem? Anyone look into it? It seems that Bill posted this a while ago and has received no reply on the list.

Signature · 3 lines
Misha Nasledov
[email protected]
GPG: A063 B99A 2BD3 2D48 F2D7 8E68 0F27 4D21 948F 8F06